- Email: [email protected]
An upper limit for the branching ratio of the decay Ks → e+e−
6 November 1997
PHYSICS LETTERS B ELSEVIER
Physics Letters B 413 (1997) 232-238
An upper limit for the branching ratio of the decay K s
e +e -
CPLEAR Collaboration A. Angelopoulos a, A. Apostolakis a, E. Aslanides k G. Backenstoss b p. Bargassa ,1 O. Behnke q, A. Benelli i, V. Bertin k F. Blanc g,m, p. Bloch d p. Carlson o M. Carroll i J. Carvalho e, E. Cawley i S. Charalambous P, G. Chardin n M.B. Chertok ¢, A. Cody i, M. Danielsson o, M. Dejardin ", J. Derre ", A. Ealet k, C. Eleftheriadis P, I. Evangelou b, L. Faravel g, R. Ferreira-Marques e, W. Fetscher q, M. Fidecaro ~ A. Filip~i~ J, D. Francis c j. Fry i, E. Gabathuler i R. Garnet i D. Garreta n, H.-J. Gerber q, A. Go n, A. Haselden i, p.j. Hayman ~, F. Henry-Couannier k, R.W. Hollander f, E. Hubert k, K. Jon-And o, P.-R. Kettle m, P. Kokkas d, R. Kreuger f,m, R. Le Gac k, F. Leimgruber b, A. Liolios P, E. Machado e, I. Mandi6 J, N. Manthos h, G. Marel n, M. Miku~ J, J. Miller c, F. Montanet k, A. Muller n, T. Nakada m, B. Pagels q, I. Papadopoulos P, P. Pavlopoulos b, j. Pinto da Cunha e, A. Policarpo e, G. Polivka b R. Rickenbach b B.L. Roberts c, T. Ruf d L. Sakeliou a, p. Sanders i, M. Sch~ifer q L.A. Schaller g T. Schietinger b, A. Schopper d, A. Soares n, L. Tauscher b, C. Thibault 1, F. Touchard k, C. Touramanis d, F. Triantis h, E. Van Beveren e, C.W.E. Van Eijk f, S. Vlachos b, p. Weber q, O. Wigger m, M. Wolter q, C. Yeche ", D. Zavrtanik J, D. Zimmerman c a University of Athens, Greece b University of Basle, Switzerland c Boston University, USA d CERN, Geneva, Switzerland LIP and University of Coimbra, Portugal f Delft University of Technology Netherlands g University of Fribourg, Switzerland h University ofloannina, Greece i University of Liverpool, UK J J. Stefan Inst. and Phys. Dep., University ofLjubljana, Slovenia k CPPM, IN2P3-CNRS et Universit£ d'Aix-Marseille 11, Marseille, France CSNSM, IN2P3-CNRS, Orsay, France m Paul Scherrer lnstitut (PSI), ViUigen, Switzerland n CEA, DSM/DAPNIA CE-Saclay, France o KTH-Stockholm, Sweden P UniversiW of Thessaloniki, Greece q ETH-IPP Ziirich, Switzerland
0370-2693/97/$17.00 © 1997 Published by Elsevier Science B.V. All rights reserved. PII S 0 3 7 0 - 2 6 9 3 ( 9 7 ) 0 1 1 8 7 - 8
A. Angelopoulos et al. / Physics Letters B 413 f 1997) 232-238
233
Received 5 September 1997 Editor: L. Montanet
Abstract was performed within the framework of the CPLEAR experiment. Full event A search for the decay Ks -+ e+ereconstruction together with e/r separation allowed powerful background rejection and high signal acceptance. The analysis of the complete set of data yields the result: BR(K, -+ e+e-) < 1.4 X lo-’ (90% CL), an improvement on the
current experimental limit by a factor of 20. 0 1997 Published by Elsevier Science B.V.
1. Introduction The decay Ks + e+e-, like the decay K, + e+eor K, + y+p-, is a flavour-changing neutral-current process, suppressed in the Standard Model and dominated by the two-photon intermediate state [l-3]. For both Ks and K, the e+echannel is much more suppressed than the ~‘JLchannel (by a factor of = 2 X 103) because of the e-p mass difference [1,2]. In terms of branching ratios there is a further suppression from K, to Ks (by a factor of = 600) resulting from the mean-life ratio T~/T,_. The branching ratio BR(K, + ewe-) is expected to be = 7 X lo-l5 [l-3]. A value significantly higher than expected would point to new physics. The current experimental limit on BR(Ks + e+e-) is < 2.8 X 10m6 (90% CL) [4]. Here we report on a new measurement of this branching ratio using the CPLEAR detector at LEAR.
with a branching ratio of 0.2% per channel. Antiprotons with a momentum of 200 MeV/c are supplied by LEAR at a rate of 1 MHz. They annihilate at rest with protons, in a pressurized hydrogen gas target, in the centre of the detector. The detector is placed in a solenoid providing a uniform magnetic field of 0.44 T. Annihilation products are detected with ten tracking-chamber layers consisting of multiwire proportional chambers, drift chambers and streamer tubes. Particle identification, in particular charged-kaon identification, is performed by a 32-sector scintillator-Cherenkov-scintillator detector (PID - Particle Identification Detector) allowing pulse-height and time-of-flight measurements. Electron and photon showers are sampled by an 18-layer electromagnetic (e.m.> calorimeter consisting of high-gain tubes of 4 X 4.5 mm2 cross-section in active layers interleaved with lead plates ( = l/3 radiation length per layer). A fast, dedicated, hardwired trigger system is tuned to an efficient selection of the annihilation channels (Eq. (1)) at the highest possible rate. A detailed description of the detector can be found in Ref. [5].
2. The experiment The CPLEAR experiment was designed to study CP violation in the neutral-kaon system using strangeness-tagged K’and K”. The experiment covering the decay time interval from 0 to 20 7s is well suited for the search for rare Ks-decays, since all Ks decay inside the detector while = 96% of the longlived component decay outside. Neutral kaons are produced in pp annihilations, pp+K-r+K”
or
K+r-i?
(1)
3. Data selection
We searched for Ks + e+e- events starting from the raw data sample used also for the study of the r+ ndecay channel [6,7], containing the e.m. calorimeter information. To maximize track reconstruction quality and in order to keep the background from the K, decays low, the search was restricted to decays before the proportional chambers, resulting in
234
A. Angelopoulos et al./ Physics Letters B 413 (1997) 232-238
an effective lifetime cut of 6 7s. Most of the raw data in this region consist of K, + rrTTfY decays. The background is obtained from a Monte Carlo (MC) simulation. Simulation data include all the main K” decay channels with two charged tracks in and the final state (r+ W, rrev, ‘TT~V, 7r+Wr”) decays are normalized to the number of rTT+ rmeasured in the time interval from 1 to 2 rs using similar criteria to Refs. [6,7]. This number is N(r+ r-1 = 5.4 x 10’. The first step of the analysis consists in the selection of a clean sample of events given by Eq. (1) above, including correct assignment of the K” decay particles. Events with four tracks, zero total charge and at least one charged kaon are selected. The probability of the kinematical lC-tit requiring a missing mass
3000
a)
equal to the K” mass at the annihilation vertex must be > 10%. Events from other annihilation channels - without a K” in the final state - are further reduced by requiring that no more than two tracks can be attached to a single vertex. The MC simulation shows that in a fraction of the events the primary (annihilation) pion is misidentified as the secondary (decay) pion having a charge opposite to that of the kaon. Such events can be recognized because the distribution of the invariant mass rn(r,,nJ of the primary pion (7~~) and the opposite-charge secondary pion (7~,) clusters at the K” mass. These events are rejected by requiring that the probability of the invariant mass rn(rP,rJ being equal to the K” mass is < 5%. In a second step, we proceed to the selection of
n
I
data
v MC background
r,l,,,,l,,,,l,,,,l’,“l”“l”“r,~,, 200 300 100
O loooo
W
-
400
=
600
500
l
700 8 IO p (MeV/c)
MC signal
2 8000 2 ; 6000 -E 4000
IO
p (MeV/c) Fig. 1. Momentum
distribution
of secondary
tracks: (a) data and MC-simulated
background,
(b) MC-simulated
signal (arbitrary
scale).
A. Angelopoulos
235
et al. / Physics Letters B 413 (1997) 232-238
the K, + e+e- decay channel by using kinematical methods and electron identification. 3.1. Kinematical selection The opening angle of the secondary tracks has to be larger than 11” in order to remove e+e- pairs from y conversions or 7~’ Dalitz decays. The optimal separation between signal and background is obtained by performing a kinematical and geometrical fit to the pp-,K*7rrKo(Ko-tefe-) hypothesis with nine constraints. The constraints require conservation of energy and momentum, the missing mass at the annihilation vertex to be equal to the K” mass, the intersection of two track-helices at the annihilation vertex and two at the decay vertex, and the K” momentum to be collinear with the line
joining the two vertices. After requiring a 9C-fit probability > 20% and all the above-mentioned selection criteria, we are left with 2745 events (data set). As a reference, the number of events selected at a similar level of the analysis as rr+ rr- decays is = 8 X lo7 in the same time interval. The MC simulation indicates that the bulk of the data at this stage is still background, hardly separable from a potential signal (see Figs. 1 and 2). This background consists of 86% rTT+ +rrTTdecays and 14% rrev decays. The contribution of the other two channels is < 0.5%. The signal acceptance relative to the background from 7r+r- decays is 1.5 X 104. The 9C-fit probability distribution for the data-set events is shown in Fig. 2 overlayed with the distribution of the MC background. Fig. 2b displays this distribution for simulated K S + e+e- events.
1500 n
B E $1000 6 d =
data
v MC background
500
111
‘0.2
0.3
III
-+=I=_+ 0.4
,,,I,,,
0.5
0.6
,,I
0.7
I,,
0.8
(,,
_
0.9
ITI
1
SC-fit probability 5000
b)
1
l
MC signal
0.7
0.8
Ll~II,~,I~,~I,,,I,,,I,,,I,,,I,,,~
0.2 Fig. 2. 9C-fit probability
0.3
distribution
0.4
0.5
for (a) data and MC-simulated
0.6
background,
0.9 1 SC-fit probability (b) MC-simulated
signal (arbitrary
scale).
236
A. Angelopoulos et d/Physics
Letters B 413 (1997) 232-238
3.2. Electron identification
recognized as electrons, that is both tracks must have a low pion-probability T(T) in the calorimeter (and be identified as electrons by the PID for momenta < 200 MeV/c). For a pion-probability cut of 4 % the dependence of the electron identification efficiency on the momentum is shown in Fig. 3. Fig. 4a gives, for the data and the expected background, the number of events that survive the kinematical selection and the electron identification, as a function of the pion-probability cut used for e/r separation. In Fig. 4b the efficiency of the electron identification is shown for MC signal events. As a result, electron identification provides an additional background rejection factor of = lo2 with an 18% probability cut and of = lo3 with a 2% probability cut, accompanied by a signal loss of 57% and 85%, respectively. For these cuts the remaining
The electron identification is the main tool that provides further reduction of the background. At high momenta (above 200 MeV/c) the identification is provided exclusively by the e.m. calorimeter [8] where, because of the high granularity, e/r separation can be performed by exploiting shower topologies. Below 200 MeV/c, the electrons are further identified using energy loss in the scintillators, number of photoelectrons per unit path length in the Cherenkov counter and time-of-flight information [9], The bulk of the data-set and MC-signal events corresponds to a high e’/emomentum (see Fig. 1). The full power of electron identification is obtained by requiring both secondary tracks to be
0.6
L
:a)
0.5 6
0.4
.i
0.3
5
0.2
.-
--0----
-w
8
0.1 0
-w
MC signal
1 I I I I I I L L I I I I I I I I I L I I I I I I I I Ii
100
200
300
400
500
600
I I I I I I I
700 p
W
0
l*I*bTr 100
n
data
v
MC background
800
(MeWc)
----=+a=p 1 I
200
,,,,I““l”“I”“I’,““‘llI 300
400
I
500
600
700
800
p (MN/c) Fig. 3. Efficiency of electron background (see text).
identification with a pion probability 9(a)
= 4% for (a) MC-simulated
signal, (b) data and MC-simulated
A. Angelopoulos
2
4
et al. /Physics
6
231
Letters B 413 (1997) 232-238
8
10
12
14
16
18
Pm
Eb)
(%I
----
o MC signal
0.1 0
1
I
2
Fig. 4. Number of data and MC-simulated probability 2%~) (see text).
background consists of K” + rear of 80% and 9.5%, respectively.
1
4 background
I
6
I
8
I
10
events (a) and efficiency
events to a level
4. Results We have compared the number of data events surviving the 9C-fit and electron identification cuts with the number of MC background events, over a wide range of confidence levels. In addition, having an exponential signal and a flat background, the upper limit of the lifetime interval was varied to search for the optimum value. Different combinations of cuts were tested by varying the 9C-fit probability cut from 20% to 90% in steps of lo%, the pion-probability cut from 2% to 18% in steps of 2% and the lifetime cut from 2 to 6 7s in 1 rs steps. In all cases, regardless of the cut combination, the
I
12
I
14
for MC-simulated
I
16
18
signal events (b) as a function of the pion
number of data events found was consistent with the number of background events expected from the MC simulation, pb. The number of data events surviving the cuts ranged from 0 (at 90% 9C-fit, 2% pion probability and 2 7s lifetime cuts) to 42 (at 20% BC-fit, 18% pion-probability and 6 rs lifetime cuts). The corresponding values for pFtb are < 0.003 and 41 f 4. The final cuts are 70% for the 9C-tit probability, 4% for the pion-probability and 3 rs for the lifetime. They correspond to the maximum significance of the potential signal, defined as v(e’ e- >/ 6, where v(e’e-) is the signal acceptance. After applying these cuts on data, no events are left, while pFLb = 0.22 f 0.10. The upper limit for the number of signal events from a Poisson-distributed process including a back-
238
A. Angelopoulos
ground is obtained using the following for the confidence level 1 - E:
e-(ru+& /Lb
+
et al./ Physics Letters B 413 (1997) 232-238
equation
[lo]
p$,n!
BR(K,
0
l--e=l-
n,
With the values reported above, Eq. (3) translates to an upper limit for the branching ratio of the decay KS + e+e-:
(2)
-+ e+e-)
< 1.4 X lo-’
(90% CL),
(4)
representing an order of magnitude improvement compared with the previous measurement [4].
0
where /_L, and n, are the number of signal events and the number of observed events, respectively. At 90% CL with no observed events and 0.2 expected (MC) background events, the upper limit for the number of signal events is p, = 2.3. Systematic uncertainties that could influence our result were tested as follows: - By shifting the number of expected background events by one sigma in each direction and repeating the optimization of cuts we obtained upper limits on pu, equal to the value given above. - Normalization and estimation of acceptances were checked by using similar electron identification and selection criteria, and the same normalization method on a sample of K” + rev events (data and MC-simulated sets). We found good agreement between data and MC simulation, proving that acceptances and normalization are well understood. We conclude that there is no significant systematic error on p,. The number of signal events (p,) is related to the branching ratios BR(e+e- ) and BR(r+ m-) by
9(e+e-> lJs= v(7T+7r-)
N(T+w) ’
BR( T+ 7T-)
x BR(e’e-),
(3) where the relative acceptance $ef e - >/ r](‘rr+ 7~- ) is 21% and $.r’ rTT-) is the acceptance for the K s + QT+r- events giving the normalization N(rf n- > = 5.4 x 10’.
Acknowledgements We would like to thank the CERN LEAR staff as well as the technical and engineering staff of our institutes for their support and co-operation. This work was supported by the following institutions: the French CNRS/Institut National de Physique Nucleaire et de Physique des Particules, the French Commissariat a 1’Energie Atomique, the Greek General Secretariat of Research and Technology, the Netherlands Foundation for Fundamental Research on Matter (FOM), the Portuguese JNICT, the Ministry of Science and Technology of the Republic of Slovenia, the Swedish Natural Science Research Council, the Swiss National Science Foundation, the UK Particle Physics and Astronomy Research Council (PPARC), and the US National Science Foundation. References [l] [Z] [3] [4] [5]
L. Sehgal, Phys. Rev. 183 (1969) 1511. L. Sehgal, Phys. Rev. D 4 (1971) 1582E. G. Buchalla, Rev. Mod. Phys. 68 (1996) 1125. A.M. Blick, Phys. Lett. B 334 (1994) 234. R. Adler, Nucl. Instrum. Meth. A 379 (1996) 76. 161R. Adler, Phys. Lett. B 363 (1995) 243. 171CPLEAR Coil.. to be published. [81R. Adler, Nucl. Instrum. Meth. A 390 (1997) 293. separation in the CPLEAR experiment [91 Electron-muon-pion using the Particle Identification Detector, CPLEAR Int. Note, January 1995. [lOI Particle Data Group, R.M. Bamett et al., Phys. Rev. D 54 (1996) 1.
PHYSICS LETTERS B ELSEVIER
Physics Letters B 413 (1997) 232-238
An upper limit for the branching ratio of the decay K s
e +e -
CPLEAR Collaboration A. Angelopoulos a, A. Apostolakis a, E. Aslanides k G. Backenstoss b p. Bargassa ,1 O. Behnke q, A. Benelli i, V. Bertin k F. Blanc g,m, p. Bloch d p. Carlson o M. Carroll i J. Carvalho e, E. Cawley i S. Charalambous P, G. Chardin n M.B. Chertok ¢, A. Cody i, M. Danielsson o, M. Dejardin ", J. Derre ", A. Ealet k, C. Eleftheriadis P, I. Evangelou b, L. Faravel g, R. Ferreira-Marques e, W. Fetscher q, M. Fidecaro ~ A. Filip~i~ J, D. Francis c j. Fry i, E. Gabathuler i R. Garnet i D. Garreta n, H.-J. Gerber q, A. Go n, A. Haselden i, p.j. Hayman ~, F. Henry-Couannier k, R.W. Hollander f, E. Hubert k, K. Jon-And o, P.-R. Kettle m, P. Kokkas d, R. Kreuger f,m, R. Le Gac k, F. Leimgruber b, A. Liolios P, E. Machado e, I. Mandi6 J, N. Manthos h, G. Marel n, M. Miku~ J, J. Miller c, F. Montanet k, A. Muller n, T. Nakada m, B. Pagels q, I. Papadopoulos P, P. Pavlopoulos b, j. Pinto da Cunha e, A. Policarpo e, G. Polivka b R. Rickenbach b B.L. Roberts c, T. Ruf d L. Sakeliou a, p. Sanders i, M. Sch~ifer q L.A. Schaller g T. Schietinger b, A. Schopper d, A. Soares n, L. Tauscher b, C. Thibault 1, F. Touchard k, C. Touramanis d, F. Triantis h, E. Van Beveren e, C.W.E. Van Eijk f, S. Vlachos b, p. Weber q, O. Wigger m, M. Wolter q, C. Yeche ", D. Zavrtanik J, D. Zimmerman c a University of Athens, Greece b University of Basle, Switzerland c Boston University, USA d CERN, Geneva, Switzerland LIP and University of Coimbra, Portugal f Delft University of Technology Netherlands g University of Fribourg, Switzerland h University ofloannina, Greece i University of Liverpool, UK J J. Stefan Inst. and Phys. Dep., University ofLjubljana, Slovenia k CPPM, IN2P3-CNRS et Universit£ d'Aix-Marseille 11, Marseille, France CSNSM, IN2P3-CNRS, Orsay, France m Paul Scherrer lnstitut (PSI), ViUigen, Switzerland n CEA, DSM/DAPNIA CE-Saclay, France o KTH-Stockholm, Sweden P UniversiW of Thessaloniki, Greece q ETH-IPP Ziirich, Switzerland
0370-2693/97/$17.00 © 1997 Published by Elsevier Science B.V. All rights reserved. PII S 0 3 7 0 - 2 6 9 3 ( 9 7 ) 0 1 1 8 7 - 8
A. Angelopoulos et al. / Physics Letters B 413 f 1997) 232-238
233
Received 5 September 1997 Editor: L. Montanet
Abstract was performed within the framework of the CPLEAR experiment. Full event A search for the decay Ks -+ e+ereconstruction together with e/r separation allowed powerful background rejection and high signal acceptance. The analysis of the complete set of data yields the result: BR(K, -+ e+e-) < 1.4 X lo-’ (90% CL), an improvement on the
current experimental limit by a factor of 20. 0 1997 Published by Elsevier Science B.V.
1. Introduction The decay Ks + e+e-, like the decay K, + e+eor K, + y+p-, is a flavour-changing neutral-current process, suppressed in the Standard Model and dominated by the two-photon intermediate state [l-3]. For both Ks and K, the e+echannel is much more suppressed than the ~‘JLchannel (by a factor of = 2 X 103) because of the e-p mass difference [1,2]. In terms of branching ratios there is a further suppression from K, to Ks (by a factor of = 600) resulting from the mean-life ratio T~/T,_. The branching ratio BR(K, + ewe-) is expected to be = 7 X lo-l5 [l-3]. A value significantly higher than expected would point to new physics. The current experimental limit on BR(Ks + e+e-) is < 2.8 X 10m6 (90% CL) [4]. Here we report on a new measurement of this branching ratio using the CPLEAR detector at LEAR.
with a branching ratio of 0.2% per channel. Antiprotons with a momentum of 200 MeV/c are supplied by LEAR at a rate of 1 MHz. They annihilate at rest with protons, in a pressurized hydrogen gas target, in the centre of the detector. The detector is placed in a solenoid providing a uniform magnetic field of 0.44 T. Annihilation products are detected with ten tracking-chamber layers consisting of multiwire proportional chambers, drift chambers and streamer tubes. Particle identification, in particular charged-kaon identification, is performed by a 32-sector scintillator-Cherenkov-scintillator detector (PID - Particle Identification Detector) allowing pulse-height and time-of-flight measurements. Electron and photon showers are sampled by an 18-layer electromagnetic (e.m.> calorimeter consisting of high-gain tubes of 4 X 4.5 mm2 cross-section in active layers interleaved with lead plates ( = l/3 radiation length per layer). A fast, dedicated, hardwired trigger system is tuned to an efficient selection of the annihilation channels (Eq. (1)) at the highest possible rate. A detailed description of the detector can be found in Ref. [5].
2. The experiment The CPLEAR experiment was designed to study CP violation in the neutral-kaon system using strangeness-tagged K’and K”. The experiment covering the decay time interval from 0 to 20 7s is well suited for the search for rare Ks-decays, since all Ks decay inside the detector while = 96% of the longlived component decay outside. Neutral kaons are produced in pp annihilations, pp+K-r+K”
or
K+r-i?
(1)
3. Data selection
We searched for Ks + e+e- events starting from the raw data sample used also for the study of the r+ ndecay channel [6,7], containing the e.m. calorimeter information. To maximize track reconstruction quality and in order to keep the background from the K, decays low, the search was restricted to decays before the proportional chambers, resulting in
234
A. Angelopoulos et al./ Physics Letters B 413 (1997) 232-238
an effective lifetime cut of 6 7s. Most of the raw data in this region consist of K, + rrTTfY decays. The background is obtained from a Monte Carlo (MC) simulation. Simulation data include all the main K” decay channels with two charged tracks in and the final state (r+ W, rrev, ‘TT~V, 7r+Wr”) decays are normalized to the number of rTT+ rmeasured in the time interval from 1 to 2 rs using similar criteria to Refs. [6,7]. This number is N(r+ r-1 = 5.4 x 10’. The first step of the analysis consists in the selection of a clean sample of events given by Eq. (1) above, including correct assignment of the K” decay particles. Events with four tracks, zero total charge and at least one charged kaon are selected. The probability of the kinematical lC-tit requiring a missing mass
3000
a)
equal to the K” mass at the annihilation vertex must be > 10%. Events from other annihilation channels - without a K” in the final state - are further reduced by requiring that no more than two tracks can be attached to a single vertex. The MC simulation shows that in a fraction of the events the primary (annihilation) pion is misidentified as the secondary (decay) pion having a charge opposite to that of the kaon. Such events can be recognized because the distribution of the invariant mass rn(r,,nJ of the primary pion (7~~) and the opposite-charge secondary pion (7~,) clusters at the K” mass. These events are rejected by requiring that the probability of the invariant mass rn(rP,rJ being equal to the K” mass is < 5%. In a second step, we proceed to the selection of
n
I
data
v MC background
r,l,,,,l,,,,l,,,,l’,“l”“l”“r,~,, 200 300 100
O loooo
W
-
400
=
600
500
l
700 8 IO p (MeV/c)
MC signal
2 8000 2 ; 6000 -E 4000
IO
p (MeV/c) Fig. 1. Momentum
distribution
of secondary
tracks: (a) data and MC-simulated
background,
(b) MC-simulated
signal (arbitrary
scale).
A. Angelopoulos
235
et al. / Physics Letters B 413 (1997) 232-238
the K, + e+e- decay channel by using kinematical methods and electron identification. 3.1. Kinematical selection The opening angle of the secondary tracks has to be larger than 11” in order to remove e+e- pairs from y conversions or 7~’ Dalitz decays. The optimal separation between signal and background is obtained by performing a kinematical and geometrical fit to the pp-,K*7rrKo(Ko-tefe-) hypothesis with nine constraints. The constraints require conservation of energy and momentum, the missing mass at the annihilation vertex to be equal to the K” mass, the intersection of two track-helices at the annihilation vertex and two at the decay vertex, and the K” momentum to be collinear with the line
joining the two vertices. After requiring a 9C-fit probability > 20% and all the above-mentioned selection criteria, we are left with 2745 events (data set). As a reference, the number of events selected at a similar level of the analysis as rr+ rr- decays is = 8 X lo7 in the same time interval. The MC simulation indicates that the bulk of the data at this stage is still background, hardly separable from a potential signal (see Figs. 1 and 2). This background consists of 86% rTT+ +rrTTdecays and 14% rrev decays. The contribution of the other two channels is < 0.5%. The signal acceptance relative to the background from 7r+r- decays is 1.5 X 104. The 9C-fit probability distribution for the data-set events is shown in Fig. 2 overlayed with the distribution of the MC background. Fig. 2b displays this distribution for simulated K S + e+e- events.
1500 n
B E $1000 6 d =
data
v MC background
500
111
‘0.2
0.3
III
-+=I=_+ 0.4
,,,I,,,
0.5
0.6
,,I
0.7
I,,
0.8
(,,
_
0.9
ITI
1
SC-fit probability 5000
b)
1
l
MC signal
0.7
0.8
Ll~II,~,I~,~I,,,I,,,I,,,I,,,I,,,~
0.2 Fig. 2. 9C-fit probability
0.3
distribution
0.4
0.5
for (a) data and MC-simulated
0.6
background,
0.9 1 SC-fit probability (b) MC-simulated
signal (arbitrary
scale).
236
A. Angelopoulos et d/Physics
Letters B 413 (1997) 232-238
3.2. Electron identification
recognized as electrons, that is both tracks must have a low pion-probability T(T) in the calorimeter (and be identified as electrons by the PID for momenta < 200 MeV/c). For a pion-probability cut of 4 % the dependence of the electron identification efficiency on the momentum is shown in Fig. 3. Fig. 4a gives, for the data and the expected background, the number of events that survive the kinematical selection and the electron identification, as a function of the pion-probability cut used for e/r separation. In Fig. 4b the efficiency of the electron identification is shown for MC signal events. As a result, electron identification provides an additional background rejection factor of = lo2 with an 18% probability cut and of = lo3 with a 2% probability cut, accompanied by a signal loss of 57% and 85%, respectively. For these cuts the remaining
The electron identification is the main tool that provides further reduction of the background. At high momenta (above 200 MeV/c) the identification is provided exclusively by the e.m. calorimeter [8] where, because of the high granularity, e/r separation can be performed by exploiting shower topologies. Below 200 MeV/c, the electrons are further identified using energy loss in the scintillators, number of photoelectrons per unit path length in the Cherenkov counter and time-of-flight information [9], The bulk of the data-set and MC-signal events corresponds to a high e’/emomentum (see Fig. 1). The full power of electron identification is obtained by requiring both secondary tracks to be
0.6
L
:a)
0.5 6
0.4
.i
0.3
5
0.2
.-
--0----
-w
8
0.1 0
-w
MC signal
1 I I I I I I L L I I I I I I I I I L I I I I I I I I Ii
100
200
300
400
500
600
I I I I I I I
700 p
W
0
l*I*bTr 100
n
data
v
MC background
800
(MeWc)
----=+a=p 1 I
200
,,,,I““l”“I”“I’,““‘llI 300
400
I
500
600
700
800
p (MN/c) Fig. 3. Efficiency of electron background (see text).
identification with a pion probability 9(a)
= 4% for (a) MC-simulated
signal, (b) data and MC-simulated
A. Angelopoulos
2
4
et al. /Physics
6
231
Letters B 413 (1997) 232-238
8
10
12
14
16
18
Pm
Eb)
(%I
----
o MC signal
0.1 0
1
I
2
Fig. 4. Number of data and MC-simulated probability 2%~) (see text).
background consists of K” + rear of 80% and 9.5%, respectively.
1
4 background
I
6
I
8
I
10
events (a) and efficiency
events to a level
4. Results We have compared the number of data events surviving the 9C-fit and electron identification cuts with the number of MC background events, over a wide range of confidence levels. In addition, having an exponential signal and a flat background, the upper limit of the lifetime interval was varied to search for the optimum value. Different combinations of cuts were tested by varying the 9C-fit probability cut from 20% to 90% in steps of lo%, the pion-probability cut from 2% to 18% in steps of 2% and the lifetime cut from 2 to 6 7s in 1 rs steps. In all cases, regardless of the cut combination, the
I
12
I
14
for MC-simulated
I
16
18
signal events (b) as a function of the pion
number of data events found was consistent with the number of background events expected from the MC simulation, pb. The number of data events surviving the cuts ranged from 0 (at 90% 9C-fit, 2% pion probability and 2 7s lifetime cuts) to 42 (at 20% BC-fit, 18% pion-probability and 6 rs lifetime cuts). The corresponding values for pFtb are < 0.003 and 41 f 4. The final cuts are 70% for the 9C-tit probability, 4% for the pion-probability and 3 rs for the lifetime. They correspond to the maximum significance of the potential signal, defined as v(e’ e- >/ 6, where v(e’e-) is the signal acceptance. After applying these cuts on data, no events are left, while pFLb = 0.22 f 0.10. The upper limit for the number of signal events from a Poisson-distributed process including a back-
238
A. Angelopoulos
ground is obtained using the following for the confidence level 1 - E:
e-(ru+& /Lb
+
et al./ Physics Letters B 413 (1997) 232-238
equation
[lo]
p$,n!
BR(K,
0
l--e=l-
n,
With the values reported above, Eq. (3) translates to an upper limit for the branching ratio of the decay KS + e+e-:
(2)
-+ e+e-)
< 1.4 X lo-’
(90% CL),
(4)
representing an order of magnitude improvement compared with the previous measurement [4].
0
where /_L, and n, are the number of signal events and the number of observed events, respectively. At 90% CL with no observed events and 0.2 expected (MC) background events, the upper limit for the number of signal events is p, = 2.3. Systematic uncertainties that could influence our result were tested as follows: - By shifting the number of expected background events by one sigma in each direction and repeating the optimization of cuts we obtained upper limits on pu, equal to the value given above. - Normalization and estimation of acceptances were checked by using similar electron identification and selection criteria, and the same normalization method on a sample of K” + rev events (data and MC-simulated sets). We found good agreement between data and MC simulation, proving that acceptances and normalization are well understood. We conclude that there is no significant systematic error on p,. The number of signal events (p,) is related to the branching ratios BR(e+e- ) and BR(r+ m-) by
9(e+e-> lJs= v(7T+7r-)
N(T+w) ’
BR( T+ 7T-)
x BR(e’e-),
(3) where the relative acceptance $ef e - >/ r](‘rr+ 7~- ) is 21% and $.r’ rTT-) is the acceptance for the K s + QT+r- events giving the normalization N(rf n- > = 5.4 x 10’.
Acknowledgements We would like to thank the CERN LEAR staff as well as the technical and engineering staff of our institutes for their support and co-operation. This work was supported by the following institutions: the French CNRS/Institut National de Physique Nucleaire et de Physique des Particules, the French Commissariat a 1’Energie Atomique, the Greek General Secretariat of Research and Technology, the Netherlands Foundation for Fundamental Research on Matter (FOM), the Portuguese JNICT, the Ministry of Science and Technology of the Republic of Slovenia, the Swedish Natural Science Research Council, the Swiss National Science Foundation, the UK Particle Physics and Astronomy Research Council (PPARC), and the US National Science Foundation. References [l] [Z] [3] [4] [5]
L. Sehgal, Phys. Rev. 183 (1969) 1511. L. Sehgal, Phys. Rev. D 4 (1971) 1582E. G. Buchalla, Rev. Mod. Phys. 68 (1996) 1125. A.M. Blick, Phys. Lett. B 334 (1994) 234. R. Adler, Nucl. Instrum. Meth. A 379 (1996) 76. 161R. Adler, Phys. Lett. B 363 (1995) 243. 171CPLEAR Coil.. to be published. [81R. Adler, Nucl. Instrum. Meth. A 390 (1997) 293. separation in the CPLEAR experiment [91 Electron-muon-pion using the Particle Identification Detector, CPLEAR Int. Note, January 1995. [lOI Particle Data Group, R.M. Bamett et al., Phys. Rev. D 54 (1996) 1.